Optimizing resiliency of distributed video surveillance system for safer city

Real-time video surveillance system is becoming an important function to keep safe city by monitoring the places attracting crowds. The system needs to be resilient so that it can timely detect abnormal events and swiftly deliver alerts to security agencies whenever events occur. In this paper, we present an architecture of resilient video surveillance system that persists even when the workload of video analysis surges by any changes in a target physical area. The proposed architecture is based on a distributed computing platform which can allocate virtual machines dynamically in response to local demand increase. In order to estimate the necessary amount of resources for video analysis, we propose a socio-ICT model that consists of a system dynamics and a queueing model. A simulation study on the model demonstrates how our platform can adapt to the changes in the target physical area and improves the resiliency.